Ensuring Non-Politically Correct AI in Federal Administration
The United States government has issued a new Executive Order on July 23, 2025, which outlines policies for large language models (LLMs) used by federal agencies. The order, known as the "Unbiased AI Principles," mandates that federal agencies procure only those LLMs consistent with two core policies: truth-seeking and ideological neutrality.
Truth-seeking requires that LLMs must be truthful, prioritize historical accuracy, scientific inquiry, and objectivity, and acknowledge uncertainty when information is incomplete or contradictory. This means that AI tools used by the federal government should provide reliable, objective information without bias.
Ideological neutrality means LLMs must be neutral, nonpartisan tools that do not manipulate responses to favor ideological dogmas, including diversity, equity, and inclusion (DEI), critical race theory, transgenderism, or related concepts. Developers are prohibited from intentionally encoding partisan or ideological judgments into outputs unless those judgments are explicitly prompted by or readily accessible to the user. The aim is to prevent what is described as "woke AI" in federal government use.
The Office of Management and Budget (OMB) has been directed to issue guidance within 120 days to assist agencies in implementing these principles. Federal contracts for procuring LLMs must include terms ensuring compliance with these principles and hold vendors accountable for costs if contracts are terminated for noncompliance. Existing contracts are to be revised accordingly as practicable, and agencies must adopt procedures to enforce compliance within 90 days after guidance issuance.
The order also allows vendors to have latitude in complying with the Unbiased AI Principles, taking different approaches to innovation. Agency heads are instructed to consider certain factors when determining whether to apply the Unbiased AI Principles to AI models other than LLMs.
This policy is aimed at ensuring that the federal government's use of AI tools is reliable, free from ideological biases, and focused on truth-seeking and objectivity. The order does not create any enforceable rights or benefits for any party, but it does require federal agencies to take steps to ensure that the AI tools they use are consistent with these principles.
- The new Executive Order in the United States emphasizes the need for large language models (LLMs) used by federal agencies to prioritize truth, prioritizing historical accuracy, scientific inquiry, and objectivity, in a policy known as truth-seeking.
- To maintain ideological neutrality, LLMs must refrain from manipulating responses to favor ideological dogmas such as diversity, equity, and inclusion (DEI), critical race theory, transgenderism, or related concepts, as specified in the Unbiased AI Principles.
- The Office of Management and Budget (OMB) is tasked with issuing guidance for agencies to implement the Unbiased AI Principles within 120 days, and federal contracts for procuring LLMs must include terms ensuring adherence to these principles.
- The policy on AI tools in federal government use seeks to ensure reliability, objectivity, and freedom from ideological biases, while providers are granted latitude for compliance with the Unbiased AI Principles in their innovative approaches.
